National newspaper duped into running GPT-4-written rage-click opinion piece
When Irish AIs are smiling
The Irish Times, one of the republic's respected broadsheet newspapers, has apologized for publishing an opinion article said to have been generated by AI software and submitted by an anonymous trickster.
Last Thursday, the publication ran the comment piece under the byline Adriana Acosta-Cortez, described as a 29-year-old healthcare administrator living in North Dublin, who argued that Irish women using fake tan represented "a form of cultural appropriation." The article sparked some debate online.
And then readers began to notice the byline photo of Acosta-Cortez looked odd.
The lighting looked unnatural and made parts of the image blurry, and her right pupil was bigger than her left one.
One day later, an actual Acosta-Cortez posted a message on Twitter criticizing the newspaper for accepting the submission.
"Genuinely sad that a once respectable news source has degraded themselves with such divisive tripe in order to generate clicks and traffic for their website. You need a better screening process than a believable Gmail address," Acosta-Cortez thundered.
It was a hoax; the person we were corresponding with was not who they claimed to be
On Sunday, the Times' editor Ruadhán Mac Cormaic said the organ had been duped into publishing a computer-generated article. "It now appears that the article and the accompanying byline photo may have been produced, at least in part, using generative AI technology. It was a hoax; the person we were corresponding with was not who they claimed to be. We had fallen victim to a deliberate and coordinated deception," Cormaic wrote in a statement.
"We don't take this lightly. It was a breach of the trust between The Irish Times and its readers, and we are genuinely sorry. The incident has highlighted a gap in our pre-publication procedures … It has also underlined one of the challenges raised by generative AI for news organisations. We, like others, will learn and adapt," he added.
There but for the grace of God goes El Reg. Arguably, the AI authorship of the piece is secondary to the fact that a newspaper ran an article by someone without verifying who actually submitted it. The piece could have been constructed by a human or computer; ultimately some editors thought it was good enough to run and did so. Still, the technology helped someone pen and byline an article they may otherwise have not bothered to pull together by themselves.
A person claiming to have submitted the piece told the Irish Independent they wanted "to stir the shit in terms of the extreme identity politics debate," and said the byline snap was created using OpenAI's DALL-E model, and that the article text was generated using GPT-4.
- Will LLMs take your job? Only if you let them
- GitHub, Microsoft, OpenAI fail to wriggle out of Copilot copyright lawsuit
- Slack adding generative AI to interact with colleagues, so you don't have to
- Meta does the 'We found baddies and crushed them' thing again – this time for AI
Editorial oversight aside, publishers are grappling with the rise of generative AI. The technology has rapidly advanced, and is now capable of producing coherent-though-bland text and realistic images. Some outlets, such as CNET and BuzzFeed, are leaning into the tech and were churning out content created by AI models months ago. Meanwhile, Wired said it will not publish stories containing text generated by AI, and The Guardian is still pondering how its policies are impacted by AI.
The Register's policy, for what it's worth, is pretty simple right now: write your own words, and don't plagiarize anyone let alone a bot; and while an AI system may be useful in suggesting information to include, don't trust it: check it against an authoritative source like you would any claim.
Also, we ensure writers are who they say they are.
Software designed to detect AI-generated text is, meanwhile, often unreliable, and it's tricky to tell when something was written by computers. Earlier this month, NewsGuard warned that "a new generation of content farms is on the way" after it found 49 websites publishing what appeared to be machine-fabricated click-bait stories making money off digital advertising. As if humans would do such a thing! ®